Re: 2.6.19-rc4: known unfixed regressions

[Alan Cox <alan@lxorguk.ukuu.org.uk>]

Ar Iau, 2006-11-02 am 12:10 -0800, ysgrifennodd Andrew Morton:
> Balls are being dropped.
> 
> > http://bugzilla.kernel.org/show_bug.cgi?id=7082
> 
> So this was a good patch but because of a bug in ne2k-pci which nobody is
> fixing we need to drop it?

I believe the patch is fundamentally wrong. We don't *need* to drop the
IO decode in this case. We don't want to drop it when the BIOS lacks the
brains to put it back. We will also kill some machines doing it as they
have devices we attach drivers to which are not just managing the
function Linux knows about but also many other things. Take the CS5520
for example, generically disable the I/O on that because we have an IDE
driver attached to it and you kill the box stone dead, as its also the
video and a few other things behind the scenes. That is not atypical.

IFF someone has a device that actually needs to disable I/O cycles, well
they can do it themselves.

Alan